Slashdot Mirror
Dot-Org TLD Signed For DNSSEC
graychase writes "A major milestone is reached as the first major top-level domain (.org) is now secured with DNSSEC. The expense to .org for implementing DNSSEC on its infrastructure and operations has not been a small one. While specific figures as to the cost of DNSSEC implementation haven't been released, Afilias, which is the technical operator of the .org registry, told InternetNews.com in 2009 that the DNSSEC implementation would be a multi-million-dollar effort. The cost isn't going to be passed on by .org to domain registrars. The move toward securing the .org registry with DNS security started in September 2008, following the Kaminsky DNS flaw disclosure."
More likely simply that different companies/organizations are responsible for .org vs .com vs .net vs .whatever, and each of those had different plans (or no plans) and acted on them at various speeds.
Because of the size of the new DNS Resource Records, notably the RRSIG and DNSKEY RRs, and partly because of the (perhaps temporarily) short TTL of one day, there will be a lot more TCP queries because of the size limit on UDP ones. The .ORG nameservers are also IPv6ified, and there is even less space in UDPv6 queries, so hosts that do not exclusively or preferentially make DNS queries in IPv4 will now make TCPv6 queries. These are likely to be slower than UDPv4 queries before the signing and v6ification, and the UDPv6 queries before the signing.
Scaling is helped by using anycast IP and IPv6 addresses, but the downside is that a routing flap that occurs any time after the first TCP/TCPv6 SYN from a client will cause a client to have to requery because of an RST fired back by the newly-closest anycast nameserver, or wait on a full TCP timeout (and then probably still see the RST) depending on the timing. (The worst case is probably having the final FIN segment being eaten by Shub-Internet or someone trying to do a devious (and probably pretty local in scope) denial-of-service consuming resources on possibly the client and two servers).
In short, this is not a win for performance, and it will be a good idea to use long TTLs in the zone itself (and on 2nd level nameservers) once it appears safe to do so.
As an end-user, is there some way for me to tell if a domain has been authenticated along the whole chain by DNSSEC?
Yes, that's actually the entire point. Your computer ("stub resolver", the library all your programs use to do DNS queries) can either (1) not care, in which case you're really no safer than with regular DNS; (2) ask your ISPs resolver whether the records were signed, in which case you're slightly safer but not very much; or (3) demand that your ISPs resolver send it all the signatures along with the actual result, in which case you're about as safe as can be (someone would have to break/steal the keys used to sign the records, in order to cause trouble).
What you as the person using the computer see, is of course dependent on the particular programs you use and what they do with the extra information that's available. Probably most don't do anything with it yet. :(
If you don't care whether the records for your domain(s) are secure, then no.
If you do want to take advantage of the new functionality, then you need to serve some extra records and give some extra data to your registrar (I think it's just the public half of your key). I imagine the exact steps to do this would vary based on who your registrar is and which DNS server you're running.